Thread 723950393

New Upcoming RTX 6090 will have two 2nm RTX 6080 chips glued together. Expected to be around 170-180 TFLOPS. 800watt power consumption.
Mrsp: 2500-3000 USD

>>723950393 (OP)
Also to have 64GB GDDR7X

op

GDDR7X to have 512bit bus performance around 2-3 TB/s

>>723950393 (OP)
oh boy, another nvidia bbq!

I can't wait until we need to plug new PCs into an industrial 240v outlet.

>two gpus glued together
>hundreds of tflops
>800 wats
>3000usd
>will stall have to upscale from 720p to hit 60fps after 2 hours of shader comp

>>723950393 (OP)
You will buy the $600 PSU and you will be happy

>>723950393 (OP)
The 5090 launched just this year. Isn't this a little small of a gap to have the details this quickly including the MSRP?

>>723950393 (OP)
>RTX 6080 chips glued together
With wood screws?

I can't wait for Chinese GPUs to enter mass production and light a fire under Nvidia and AMDs ass. I'm buying a LiSuan 7G106 as soon as they're available, which will hopefully be in March.

Whats the point of tech like this if you arent working on AI on your personal computer

>>723951045
i'll definitely buy a chingaling gpu during it's third or fourth generation. the first generation will be literal bombs.

>>723950393 (OP)
Aren't the 80 grade chips just a cut down 90 chip?
Wouldn't having 2 GPU's linked together create all kinds of bottlenecks and other inefficiencies?
How does a jump from 5nm to 2nm result in each 5080 chip, the most powerful single chip that Nvidia will manufacture for the gen being less powerful than a single 5090 grade chip?
None of this makes any sense.

>>723951045
This shit better be like $100 because it's complete shit
>6nm
>225w
>no rt
>24 tflops
On paper it's similar to an RTX 3070 but it gets the shit kicked out of it by an RTX 4060 which is weaker than an RTX 3070, by all metrics this GPU is a complete pile of shit.
>expected price $300 to $400
You're literally better off buying an AMD,Nvidia or Intel equivalent GPU, they will consume less power, be priced either less or the same, and perform better on all games.

>>723952598
And only Chinese games devs will support it probably

>>723950910
silence goyim

>>723952678
Most likely, it already runs Black Myth Wukong has already been optimized for it, maybe the person I replied to is really into Chinese games, and Chinese mythology, and Chinese products and Chinese politics, and Chinese genetics.

>>723950681
Industrial outlets are 360V.
But think about burgers they can only pull 1500W because of 120V, kek. Or they will need separate 240V powerline.

>>723950393 (OP)
will this card be enough for glorious UE6 rays in 720p?

>>723950808
Yes thats the idea basically instead of doing the sli bullshit with GTX690.
We are gonna just glue two dies together

>>723951679
Not really.
Yes it would be a 800watt BBQ so you need a 1200W PSU minimum with two 12VPWR connectors

Let me guess it's gonna burn some houses down

>>723953278
Nah. But you might get above 30 fps in Cyberpunk now.

I regret not buying a 4090 when I had the chance

rip

>>723950393 (OP)
>implyingificating the main difference won't just be how many upscalerino AI cores they stuff in it
lol, lmao even

>>723950393 (OP)
SLI cards are trash. it never worked properly

It will take 4 months after launch for sales to get near to MSRP anyway.

Past the honeymoon phase, where youtubers+plus forum shills have stopped being paid to glaze it and we'll know just how shit it is.

>>723950956
To be able to withstand the heat, male glue will be needed periodically.

>>723952598
Lisuan demonstrated the GPU running Wukong at 1440P/High at around 70FPS. If they choose to undercut the competition on price at sub-$250, I wouldn't mind paying the import duties. It just has to reach Arc B580 level performance, but without any of Intel's annoying ReBar requirements, x8 Pcie lanes, and high CPU overhead.

>>723950393 (OP)
>800watt power consumption
dual PSU cases when?

Can we just have a separate wall plug for gpus

>>723950393 (OP)
for what games doe?

>>723958932
do you not play games?

>>723950393 (OP)
If it doesn't have at least 64gb of VRAM it's SHIT. At that price it should be 128gb.

>>723951679
It depends on how they are linked together. China has been stuck with duv lithography (late 90's tech) so they have had to develop a bunch of workarounds to catch up to ASML. they are getting close and this dual-die connectivity is part of that.

>>723950808
You can always spot the poors because they think 60fps is the target framerate for high end cards.

>>723950393 (OP)
>buying an AIslop generator and then playing games on it

>>723950393 (OP)

Still won´t be able to run the latest AAA slop at consistent 60 FPS though.

>double the price
>double the size
>double the power consumption
>WOAH LOOK AT THESE PERFORMANCE GAINS
They've been doing this for like 3 generations now. This is not innovation. I hate Nvidia

>>723962618
sneak peak of the 7090 bro! it's $10,000 and gonna require 2 seperate outlets to power but you're gonna get 30% more FPS in the new DOOM (dlss on)

>>723962189
In reality the target is 30 because you can use AI to add in the other 30 frames to reach 60.

I can't think of a reason to upgrade from my 6700xt.

>>723962854
In reality the target is 30 FPS because you consolefags ruin everything with your poorfag console hardware. You've ruined arcade games and you've ruined PC vidya for 12 years till the raytracing and AI upscaling told you to fuck off and buy a PC if you want to play.

>>723962854
this unironically

>>723963143
>It will need two of this connector to load balance
problem solved

>>723963509
At this point they should just power GPUs directly from the mains.

>>723958801
>internal demo
Yawn, they always pull this shit and then the final product is always a pile of shit, I'm surprised anyone still believes any of this shit whenever it has china attached to it, bro have you seen those bolt graphics gpus.